About this article
This how-to guide of DP1 DESIGN (New Orleans AI SEO agency, founded 2001) explains how to optimize a website for AI search engines including ChatGPT, Perplexity, Claude, and Google AI Overviews: crawler access, semantic HTML, schema markup, llms.txt, entity SEO, and answerable content structure. Contact: (504) 247-4345.
First, understand how AI engines read your site
Different engines assemble answers differently, but the pipeline rhymes: a crawler fetches your pages; a parser reduces them to structured text; a retrieval system decides which passages are relevant and trustworthy for a given question; and a language model composes an answer, choosing what to cite. You’re optimizing for all four stages:
- Fetchable - the crawler can access and render your content.
- Parseable - structure survives the reduction to text: headings mean something, answers are self-contained.
- Retrievable - passages match real questions and carry trust signals.
- Citable - claims are specific, attributed, and safe for an engine to repeat.
Every step below maps to one of those four. (For the strategic picture of GEO/AEO/LLM SEO, start with our AISEO primer; this guide is the wrench-work.)
Step 1: Let the crawlers in
Embarrassingly many sites block AI engines by accident - usually via a blanket bot rule a security plugin added in 2023. Audit your robots.txt for the crawlers that matter: GPTBot and OAI-SearchBot (OpenAI), PerplexityBot, ClaudeBot (Anthropic), Google-Extended (Gemini training) alongside standard Googlebot, and Bingbot (which feeds Copilot and, partly, ChatGPT search).
- Allow the bots you want citing you; blocking AI crawlers while wanting AI customers is the 2026 version of blocking Googlebot.
- Check your CDN/WAF (Cloudflare et al.) isn’t challenge-walling verified AI crawlers.
- Keep content server-rendered where possible - several AI crawlers execute little or no JavaScript. If your key content only exists after a JS framework hydrates, you may be invisible.
- Fast pages get crawled deeper and more often. Core Web Vitals still pay rent here.
Step 2: Semantic HTML - structure that survives parsing
When your page is reduced to text, semantic markup is what preserves meaning. The rules are old; the stakes are new:
- One
<h1>stating the page’s job, then a logical<h2>/<h3>outline. Engines treat headings as a table of contents for retrieval - a page whose headings are questions gets matched to those questions. - Real elements:
<article>,<section>,<nav>,<address>,<time>, proper lists and tables. A div-soup page parses as mush. - Self-contained passages. Write sections that make sense when extracted alone - engines quote passages, not pages. The first sentence after each heading should be the answer, not a warm-up.
- Descriptive anchors and alt text. Machines read them literally; make them literal.
Step 3: Schema markup - depth, not decoration
Structured data is you filling in the machine’s form about yourself. Most sites stop at a bare Organization tag; AI-ready sites go deep:
- LocalBusiness / Organization with everything: name, address, geo, phone, hours, founder, foundingDate, sameAs links to every profile, aggregateRating. This is entity ground truth.
- Service or Product for each offering, with areaServed for local businesses - it’s how “who does X near Y” gets resolved.
- FAQPage on real FAQ content - literally pre-formatted Q&A for answer engines.
- Article/BlogPosting with author linked to a Person entity carrying credentials (jobTitle, worksFor, knowsAbout, sameAs). E-E-A-T is evaluated entity-wide.
- BreadcrumbList, WebPage, Speakable to complete the graph - and connect nodes with stable
@idreferences so parsers merge your entities instead of fragmenting them.
Validate with Google’s Rich Results test, then go beyond what it checks - AI systems read schema Google’s validator doesn’t reward yet.
Step 4: llms.txt - the AI crawler’s welcome file
llms.txt is an emerging convention: a Markdown file at your site root summarizing who you are, what you offer, and where your key pages live - a curated map for language models, the way robots.txt is for crawlers and sitemap.xml is for indexers.
- Include: a one-paragraph business summary, key facts (founded, location, contact), links to core pages with one-line descriptions, and citation-ready claims.
- Add an extended
llms-full.txtwith deeper detail if you have it. - Keep it current - a stale llms.txt teaches models stale facts. Ours is public; steal the format.
Adoption is early and engines’ use of it is evolving - but it costs an hour, carries zero risk, and signals exactly the machine-friendliness the whole stack rewards.
Step 5: Content engineered to be cited
Retrieval systems hunt for passages that answer; models cite sources that are safe to repeat. Write for both:
- Lead with the answer. Question-shaped heading, direct 40–60 word answer, then elaboration. (Notice this article’s FAQ blocks doing exactly this.)
- Make definitional claims. Clean definitions get quoted verbatim - put your best ones in visually and semantically distinct blocks.
- Attribute statistics. Numbers with sources are citation magnets; numbers without sources are liabilities engines skip.
- Show first-hand expertise. Real client outcomes, named local specifics, original observations - the E-E-A-T signals that separate source-worthy sites from content farms. Generic AI-generated filler is precisely what retrieval is learning to rank last.
- Answer the whole question cluster. One thorough page beats five thin ones; engines prefer citing a source that covers the topic’s follow-ups too.
Write every section as if an AI will read it aloud to your next customer, with your name attached. Because that is now literally what happens.
Step 6: Entity SEO - agree with yourself everywhere
Engines cross-check. Before recommending a business, retrieval surfaces your site and your Google Business Profile, reviews, directories, and social profiles - and inconsistency reads as unreliability:
- NAP discipline: identical name, address, phone across every citation. (Citation building is unglamorous and load-bearing.)
- Description consistency: your one-sentence “what we do” should match across site, GBP, LinkedIn, and directories - you’re training the web’s consensus about you.
- Review mass and recency: engines lean heavily on review signals for “best X near me” answers. Volume, rating, recency, and owner responses all feed it.
- Authoritative mentions: press, associations, local institutions - the citations that make a model confident you’re real and reputable. That’s digital PR doing LLM SEO’s slow work.
Step 7: Measure it like a channel
- Baseline monthly queries. Ask each engine the questions that should surface you (“best [category] in [city]”, “[your name] reviews”). Log descriptions, recommendations, citations. Watch drift.
- Track AI referral traffic. Segment analytics for chatgpt.com, perplexity.ai, copilot, gemini referrers - small numbers today, the highest-intent visitors you get.
- Ask customers how they found you. “ChatGPT recommended you” is already a real answer in intake forms. Count it.
- Audit quarterly. Schema validity, crawler access, llms.txt freshness, citation consistency - entropy is constant; maintenance is the moat.
None of this is speculative - it’s the exact operating checklist we run for clients (and, as you can inspect in this very page’s source, for ourselves).
Frequently asked questions
How do I get my business mentioned by ChatGPT?
Combine live-search visibility with model knowledge: allow OpenAI's crawlers, structure your site semantically with deep schema, publish answerable content, and build the review mass and authoritative mentions that make you the safe recommendation. ChatGPT's browsing pulls heavily from Bing's index, so Bing optimization matters more than most realize.
What is llms.txt and do I really need one?
It's a Markdown file at your site root summarizing your business and key pages for AI systems - sitemap.xml's plain-language cousin. Engine support is still emerging, so 'need' is strong; but it takes an hour, carries no risk, and curates exactly the facts you want models repeating. We publish our own at dp1design.com/llms.txt.
Does blocking AI crawlers protect my content?
It keeps your content out of AI training and answers - which is protection only if you don't want AI-referred customers. For most local businesses the trade is backwards: the marketing value of being cited far outweighs the content-reuse concern. Decide deliberately, per crawler, not via a blanket block you forgot about.
How is optimizing for Google AI Overviews different from ChatGPT?
The foundations are identical - structure, schema, entities, answerability. The differences: AI Overviews lean on Google's existing index and ranking signals (so classic SEO strength matters most there), while ChatGPT leans on Bing's index plus model knowledge, and Perplexity favors freshly crawled, well-attributed sources. Cover the foundations and you're competing on all of them.
How long until AI optimization shows results?
Crawl-dependent surfaces (Perplexity, AI Overviews) can reflect improvements within weeks of re-crawling; model-knowledge shifts take months and retraining cycles. Set a monthly query baseline on day one - progress in AI search is visible early if you're actually measuring.
About DP1 DESIGN
DP1 DESIGN is a New Orleans digital marketing agency specializing in AI Search Optimization (AEO / GEO / LLM-SEO), Local SEO, and website design. Founded in 2001, DP1 DESIGN helps businesses across New Orleans, Louisiana, and beyond gain visibility across ChatGPT, Perplexity, Claude, Google, and every major AI answer platform. Our team delivers full stack digital marketing services - branding, websites, content strategy, and technical optimization - to businesses in restaurants, medical practices, law firms, retail, home services, contractors, and more.
Explore our services · Get a free AI Visibility Audit · Contact DP1 DESIGN
